Idea
An AI platform using street view imagery to monitor cycling and motorcycling travel behavior for urban planners and transport agencies
Research Paper
Core Innovation
This paper fine-tunes a YOLOv4 model to detect cycles and motorcycles from Google Street View images with high precision. It then applies beta regression models to predict travel mode shares accurately across diverse global cities. This method enables scalable, low-cost travel behavior monitoring where traditional survey data is unavailable or outdated.
Market Size (TAM)
$2–10B TAM, $1–2B SAM; assumption: global urban mobility analytics and smart city planning markets expanding with demand for AI-driven data solutions.
Potential Customers & Pain Points
- Urban Planners Needing Travel Mode Data
- Transport Agencies Lacking Recent Survey Data
- City Governments Seeking Cost-Effective Mobility Insights
Business Model
Subscription-based API and analytics platform offering travel mode detection and prediction services to urban planners and transport agencies
Competitive Landscape
- StreetLight Data
- INRIX
- Moovit
Implementation Challenges
- Access to up-to-date street imagery
- Variability in image quality across regions
- Integration with existing urban planning tools
Validation Strategy
- Pilot deployment with select city governments
- Compare predictions against recent travel surveys
- Iterate model based on feedback and accuracy metrics
Research Paper Overview
Vehicle detection from GSV imagery: Predicting travel behaviour for cycling and motorcycling using Computer Vision
Summary
This study uses deep learning on Google Street View images to estimate cycling and motorcycling travel behavior across 185 global cities. Using a fine-tuned YOLOv4 model, it detects cycles and motorcycles with 89% precision. Beta regression models then predict mode shares with over 61% R2 accuracy, enabling travel behavior insights in cities lacking recent survey data. The approach offers scalable, cost-effective travel mode monitoring to complement traditional data sources.